How Tavily Got to 1M Users Without Sending One Cold Email

Today, the buyer is not the person. The buyer is the AI system the person is using. We created a category. It didn't exist before. I don't want to see any competitors next to us in the next two years. We're trying to give you days of research in a second. People are going to be mad that I'm saying that.

Welcome back to the Nevis for Serves podcast. My name is Marquesha Patrice and I'll be your host today. With me is Rotim Weiss. He's the founder and CEO of Tivili, the search layer for the agentic web. Before we even, I guess, talk a little bit about Tivili, I want to get to know you. Of course. So Rotim, can you tell me a little bit about yourself and kind of how you got started on your founder's journey?

I think when it started, it was unique, right? I was a student. It's not like someone who is in the workforce already and seeing what's happening. I was still a student. ChatGPT came out and immediately I had a feeling that I got to do something right. It was the sense of revolution happening. At that time, I was reached out by Yuval, someone I knew from years back. He was a managing director at the hedge fund, and he wanted me to come join the fund. He wanted me to come and explore what the hell is happening, like what's LLM, what we need to do, where we need to invest. And I actually told him no, because I... I was like, nah, I got to start something. And he said, sounds great. Just join me. And if you find good enough of a problem, I'm going to give you the first check and hit the road. So that's exactly what I did. The student cannot say no to that. What we noticed was everyone are building this type of research agents, right? AI and computers in general has the capacity to just process massive amount of information. And they immediately utilize LLMs to do exactly that, summarize this amount of data and searching this massive amount of organizational context. But what was missing is really the real-time data. So if you remember up until a year ago, you would ask a GPT, hey, what's the weather today? And you would get the response, I'm sorry, my knowledge is cut off to 2021, they cannot answer that. What we understood, yes, the models are static. You train one time on a finite amount of data, And they get stuck. They don't know what happened in the world after that. So that was an easy solution. I just went to the web. I copy-pasted the weather today. And I asked it again, hey, what's the weather today? And it answered. We immediately understood we've got to connect the web, the real-time knowledge. to these type of agents. And it was really, really early on. And we actually opened, we launched GPT Researcher, which was an open source, skyrocketed. Like it was so needed. It got, I think, almost 30K GitHub stars pretty fast. And what it is was, today is very simple, but back in the days, you didn't see that. It just went to the web based on a query. did some research, processed a lot of data, compressed it, and gave the user a report using real-time information. We immediately understood we have a good problem to solve, and we launched the company.

And you mentioned you kind of talked to other companies. How much of your process was actually